toumpis, athanasios and galloway, alexander and cater
TRANSCRIPT
Development of a Process Envelope for Friction Stir
Welding of DH36 Steel – A Step Change
Athanasios Toumpisa*, Alexander Gallowaya, Stephen Caterb, Norman
McPhersonc
aDepartment of Mechanical & Aerospace Engineering, University of Strathclyde, James Weir
Building, 75 Montrose Street, Glasgow G1 1XJ, United Kingdom bFriction and Forge Processes Department, Joining Technologies Group, TWI Technology
Centre (Yorkshire), Advanced Manufacturing Park, Wallis Way, Catcliffe, Rotherham S60 5TZ,
United Kingdom cBAE Systems Naval Ships, 1048 Govan Road, Glasgow G51 4XP, United Kingdom
ABSTRACT
Friction stir welding of steel presents an array of advantages across many
industrial sectors compared to conventional fusion welding techniques.
However, the fundamental knowledge of the friction stir welding process in
relation to steel remains relatively limited. A microstructure and property
evaluation of friction stir welded low alloy steel grade DH36 plate, commonly
used in ship and marine applications has been undertaken. In this
comprehensive study, plates of 2000 x 200 x 6 mm were butt welded together
at varying rotational and traverse speeds. Samples were examined
microscopically and by transverse tensile tests. In addition, the work was
complemented by Charpy impact testing and micro-hardness testing in various
regions of the weld. The study examined a wide range of process parameters;
from this, a preliminary process parameter envelope has been developed and
initial process parameter sets established that produce commercially attractive
excellent quality welds through a substantial increase in the conventionally
recognised weld traverse speed.
Keywords: Friction stir welding; Optimisation; DH36.
* Corresponding author. Tel.: +44 (0)7513 621 906,
e-mail address: [email protected]
1. Introduction
Friction stir welding (FSW), invented by Wayne Thomas [1] at TWI in December
1991, is a solid state joining process in which a constantly rotating, cylindrical-
shouldered tool with a profiled probe is traversed at a constant rate along the
joint between two clamped pieces of butted material. The probe is slightly
shorter than the weld depth required, with the tool shoulder riding along the top
of the work piece surface. The material is thermo-mechanically worked and
heated high enough for plastic deformation to occur but well below its melting
point. The basic concept of the process is shown in Figure 1.
Fig. 1. The basic concept of friction stir welding.
Frictional heat is generated between the tool and the work pieces. This heat,
along with that produced by the mechanical mixing process and the adiabatic
shearing within the material causes the stirred materials to soften without
melting. As the tool is moved forward, a special profile on the probe forces
plasticised material to the rear where clamping force assists in a forged
consolidation of the weld. This process of the tool traversing along the weld line
in a plasticised tubular shaft of metal results in severe solid state deformation
involving dynamic recrystallization of the base material. FSW is a very complex
multi-physics process incorporating mechanical and thermal processing of the
material, considerable plastic deformation and high levels of flow stress; it is a
process analogous to forging rather than casting which more closely resembles
the conditions observed during conventional fusion welding.
FSW is currently being extensively employed in aluminium joining applications
but there is significant interest by many industrial sectors in transferring the
process and its advantages to steel. Preliminary studies have demonstrated the
feasibility of FSW of steel [2,3], while others have shown that there are several
positive effects on the properties of friction stir welded steel plates such as
considerable grain refinement, excellent fatigue properties and minimised
distortion [4].
The current study is focusing on FSW of steel grade DH36, a low alloy steel
utilised in the European shipbuilding industry among other sectors. For FSW of
steel to become economically and technically viable for introduction in the
shipbuilding industry, it should evolve into a process competitive with
conventional fusion welding methods. In the shipbuilding sector, this
requirement is translated into high welding speeds (in millimetres per minute)
which produce welded joints of acceptable quality. Therefore, an optimisation
study is a fundamental step towards this direction; that is, a study concerned
with establishing the limits of the process (the process envelope) in terms of the
two more significant parameters which can be directly controlled, tool traverse
speed and tool rotational speed.
There are limited studies thus far for FSW of structural steel in the relevant
technical literature, particularly for high welding speeds. In a well cited study
from 2003, Reynolds et al. [3] examine friction stir single sided welds of hot
rolled, 6.4 mm thick DH36 steel, produced by four different welding speeds in
an inert gas environment, to assess the relationship between varying weld
parameters and resultant weld properties. They [3] observe a bainitic and
martensitic microstructure in the bulk of the thermo-mechanically affected zone
(weld nugget) of the fast weld (450 mm/min). However, only this weld’s
microstructural features are reported therefore no comparison can be made to
the intermediate and slower welds. The hardness of all welds demonstrates a
continuous increase from parent material to nugget, with a variation of
approximately 190 HV up to the peak hardness of the fast weld. The tensile
tests reveal significant overmatching of all welds; longitudinal tensile tests show
that the yield strength of all welds is higher than the parent material’s UTS, and
this is attributed to the weld nugget microstructure being very different from the
original ferrite / pearlite microstructure. In all, weld hardness and strength is
seen to increase with increasing welding speed. The effect of increasing
rotational speed on weld properties is not considered in this study [3].
Further work [4] on the same grade of steel, DH36, extends the mechanical
properties assessment of friction stir welded plates, also examining three
different thicknesses (4, 6, and 8 mm) and comparing to submerged arc welded
(SAW) plates of same thickness, in order to evaluate the potential of FSW as a
shipbuilding welding process. McPherson et al. [4] observe an acicular shaped
ferrite microstructure in the thermo-mechanically affected zone, consistent over
the mid-thickness of all weld samples, and a finer unspecified structure
seemingly increasing with decreasing plate thickness. The parent material is
seen to consist of bands of ferrite and pearlite, as expected for rolled steel
plates. Variations in hardness distribution are considered minor and certainly
not expected to produce adverse effects. Likewise, impact toughness levels for
FSW and SAW samples at -20oC are reported to be similar and within
classification society impact requirements. The focus of this study [4] is shifted
towards the previously mentioned mechanical properties therefore the welds’
tensile behaviour is not discussed in detail; all transverse tensile samples
however fractured in the parent material. In conclusion, this study [4] supports
sufficiently the argument for the capability of FSW to match shipbuilding
requirements.
Ghosh et al. [5] study the friction stir lap welding of thin plates of high strength
martensitic M190 steel by optical microscopy, tensile lap shear testing and
micro-hardness testing. Their objective is to optimise the FSW of M190 steel in
the automotive industry by assessing the resultant microstructure and
mechanical properties produced by ten different sets of welding parameters,
600-1200 rpm rotational speed and 51-203 mm/min traverse speed, also
applying forced air cooling. The researchers [5] find that with increasing
traverse and rotational speed, the microstructure of the weld nugget becomes
predominantly martensitic due to very high cooling rate. Bainite is seen to
appear and gradually increase with the same rotational speed but slower
traverse speeds, hence marginally lower cooling rates. They conclude that the
temperature of the inner heat affected zone reached above the steel’s A1
temperature during welding thus austenite transformed into ferrite and pearlite
during cooling. Still, the outer heat affected zone temperature remains below A1
therefore exhibiting tempered martensite because of the martensitic parent
material and the amount of heat dissipating through this region. The results of
the tensile shear tests show that all samples fractured in the vicinity of the inner
heat affected zone that seems to be the weakest region due to its ferrite-
pearlite microstructure. Ghosh et al. [5] however do not proceed in developing
new sets of parameters which will improve the mechanical properties of the
weld zone, consequently weakening their case for an optimisation study.
In an investigation into the possible use of FSW as an alternative to electric
resistance welding of API X100 grade high strength linepipe steel, Cho et al. [6]
examine its microstructural evolution during friction stir butt welding after
significant prior preparation of the plates. Their work is mainly focused on the
grain structure development in a single set of process parameters, 127 mm/min
and 450 rpm, using optical microscopy, EBSD-SEM and TEM. Starting from the
parent material microstructure of dual phase ferrite and bainite, the thermo-
mechanically affected zone is reported as having a very fine, homogeneous
microstructure while the stir zone is acicular shaped bainitic ferrite rich. The
former is attributed to continuous dynamic recrystallization, while the latter
occurs because of the particular to this region austenite to ferrite phase
transformation under high cooling rate and high strain. Possible heterogeneity
of the weld is not discussed however as the microstructural assessment is
concentrating only on the retreating side of the weld zone. The micro-hardness
measurements [6] show that stir zone hardness is significantly higher than all
other regions, mainly due to its developed microstructure. No other mechanical
properties of the weld are presented.
There are many relevant optimisation studies for FSW of aluminium alloys
including modelling, statistical analysis or experimental work, but they seem
rather small scale projects when compared to the present work, and therefore
cannot be transferred into FSW of steel. For example, Kumar et al. [7] attempt
to optimise the FSW of two dissimilar aluminium alloys with regard to tensile
properties and hardness of the weld, with no commentary on the
microstructural evolution of the thermo-mechanically affected zone. They [7]
also explore the effect of each parameter (rotational speed, tool tilt and type of
tool pin) on these properties by percentage contribution, as derived from their
statistical analysis. Kumar et al. [7] employ the Taguchi method to direct their
analysis but this approach leads them to experimentally examine only nine sets
of parameters; in fact, traverse speed, although highly significant, is not one of
these parameters. Hence their conclusions are vague and limited.
Ghosh et al. [8] investigate the FSW process on two dissimilar aluminium alloys
for optimised strength and formability by developing appropriate sets of
parameters (rotational speed and traverse speed). They conclude [8] that
components welded at the lowest rotational and traverse speed possess
improved mechanical properties, more refined grains and decreased residual
stresses. In their study, the researchers [8] use the findings of previous studies
[9,10] for FSW of each of the two alloys individually. They then proceed in
selecting only four sets of parameters to optimise the welding of both alloys but
without further discussing how these parameters were chosen. In addition, the
effect of changing the position of the two alloys, from left to right before
welding, on the resultant microstructure and mechanical properties is not
reported.
In the present study and in contrast to previous process parameter
development studies in aluminium alloys, 19 friction stir welds, the majority of
which were produced with welding speeds not previously identified by other
research groups, were extensively examined in order to determine a step
change improvement to the current industrially accepted welding speeds.
These were selected from a comprehensive and considerable spread of FSW
DH36 butt joints in which slow, intermediate and fast welding speeds were
trialled. This extensive work, the first of its kind on steel, has led to an initial
understanding towards developing the fundamental knowledge on friction stir
welding of steel.
2. Development of process parameter envelope
The process parameter envelope for FSW of DH36 steel can be represented by
plotting the two more important weld parameters, tool rotational speed and
traverse speed as variables on the x and y axes of a graph (Figure 2). The
development of the envelope begins with a set of parameters that are known to
deliver acceptable quality welds on steel. It is then broadened outwards from
these parameters, with varying rotational and traverse speeds, in order to
establish the outer limits of the process envelope by producing and examining a
series of welds. The envelope consists of the parameter sets that produce
welds of at least acceptable quality, and is divided in this case into three groups
characterised by welding (traverse) speed. The shape of this envelope is
expected to vary with any further development and refinement of the process
parameters. To generate the process envelope, extensive welding and
considerable testing is required over a very large data set.
Fig. 2. Schematic representation of the process parameter envelope.
The process parameter envelope development in the present study
commenced using the process parameters recommended by the FSW tool
manufacturer, MegaStir Inc. of Provo, Utah, USA. These are based upon a tool
traverse speed of 100 mm/min and a tool rotational speed of 200 rpm. In
simplistic terms, the tool rotational speed controls the frictional heat input to the
weld zone and plasticises the steel whilst the tool traverse speed influences the
weld cooling rate and thus has a significant effect upon the evolving
microstructure. Other factors, for example the tool tilt, tool plunge depth, tool
cooling, backing bar characteristics and clamping arrangements have a
secondary but not insignificant effect upon the welding process.
The process parameters used cannot be considered in isolation from the tool
itself. Whilst it is desirable to use process parameters that produce the best
weld in terms of microstructure and mechanical properties, the current tool
technology for FSW of steel is still relatively immature and attention must be
paid to ensure that the welding environment is not so aggressive that the tool’s
life is shortened by the welding process. A compromise must therefore be
made between selecting parameters that deliver good weld properties and
parameters that extend the life of the tool and thereby improve the economic
viability of the FSW process in steel.
In the present study, a number of welds were initially produced using the
previously stated process parameters as this provided a baseline data set of
the generated forces, torques and heat inputs resulting from these parameters.
These data framed the basis for expanding the process envelope to higher
welding speeds.
It is desirable that FSW is competitive to conventional fusion welding
techniques in terms of welding speed, hence the process parameters were
extended to explore the potential to use higher welding traverse speeds. Thus,
if the tool is to be traversed through the steel being welded more quickly, then it
is necessary to establish that the steel ahead of the tool is sufficiently
plasticised to ensure that the forces experienced by the tool do not increase to
the point where the tool fails, and the steel continues to flow around the tool
and can be consolidated behind it effectively.
To weld at higher traverse speeds, the tool rotational speed needs to be
increased. Increasing the tool rotational speed increases the heat input to the
weld, but does not do so in a linear manner; above a certain point, the friction
couple between the steel and the tool changes and the heat input into the weld
can decrease. Parameter selection is therefore a complex process with many
interdependent variables, many of which are currently poorly understood.
Once a baseline understanding of the welding process was achieved, the
primary process parameters of tool rotational speed and tool traverse speed
were increased. The objective was to increase the welding speed from 100
mm/min to 350 mm/min, the point at which FSW would become competitive to
fusion welding processes in terms of production speed of acceptable quality.
Further increases in welding speed beyond 350 mm/min are desirable to
improve the competitiveness of FSW, thus faster welding speeds were also
investigated. Table 1 shows the welding speeds investigated in this initial
process parameter development work. The heat input calculations are based
on the equation presented by McPherson et al. [4] such that: H=ε2πRτ/1000V,
where ε dimensionless factor of FSW process efficiency, R rotational speed in
rpm, τ average steady state torque on the tool in Nm, and V traverse speed in
mm/min.
Table 1
List of weld parameters examined.
Reference number
Rotational Speed (rpm)
Traverse Speed (mm/min)
Heat Input (kJ/mm)
W019A 200 100 3.44
SL
OW
W019B 200 110 3.07
W019C 200 120 2.68
W020A 200 130 2.61
W020B 200 143 2.31
W020C 200 156 2.04
W021 400 200 3.07
W022 400 250 2.29
INT
ER
ME
DIA
TE
W023 400 275 2.13
W027 400 325 1.77
W028 400 375 1.45
W029 450 400 1.61
W030 450 350 1.97
W031 300 250 1.92
W032 600 500 1.77
FA
ST
W033 650 500 2.12
W034 575 500 1.35
W035 700 500 2.23
W036 675 500 2.01
Through this process parameter development, the state of the art has been
increased from conventionally adopted welding speed of 100 mm/min to a more
commercially attractive speed of 400 mm/min. Although a step change in the
welding speed has been identified, the purpose of the present study is to
assess the impact of this increase on the microstructural evolution and
mechanical properties of each friction stir weld.
3. Experimental procedures
3.1. Material and welding details
A substantial number of single sided friction stir butt welds in 6mm thick DH36
steel were produced using a PowerStir FSW machine operated in position
control. All plates were 2000 mm in length and 200 mm in width, therefore
fabricating a welded component of 400 mm in width. The rolled plates were
welded in the as received condition and were heavily clamped to the machine
bed, without any prior surface preparation. The nominal chemical composition
of the shipbuilding steel grade DH36 which was examined in this study is
presented in Table 2, as provided by the steel supplier.
Table 2
Chemical composition of 6 mm thick DH36 steel (wt%).
C Si Mn P S Al Nb N
0.11 0.37 1.48 0.014 0.004 0.02 0.02 0.002
Friction stir welding was performed using the current leading FSW tool
technology for steel, this being the WRe-pcBN Q70 stepped spiral tool to
design M44837 (Figure 3). This tool is equipped with a scrolled shoulder,
having a probe length of 5.7 mm and rotating anti-clockwise. The tool is
protected by an inert gas environment against oxidation due to the high
temperatures of the FSW process.
Fig. 3. Basic dimensions of the FSW tool employed for this study.
3.2. Microstructure evaluation
Microstructural characterisation was undertaken towards establishing the limits
of the FSW process. One outcome of this microstructural assessment is
generating important information regarding the mechanical properties that are
likely to be attained in the weld zone. A secondary outcome is providing
information related to undesirable process induced defects or flaws that could
compromise the integrity of the weld, the absence of which provides
reassurance that the weld process parameters reported in Table 1 will lead to
an acceptable level of quality. The resultant microstructure and its possible
variations throughout the weld region are thoroughly examined and reported
herein.
One metallography sample was transversely sectioned from a random position
within the steady state condition region of each of the nineteen welds. In the
steady state region, the forces that the FSW tool sustains stabilise (Figure 4),
the weld quality is improved, and any sample removed from any position within
the steady state region is expected to be representative of the properties of the
entire weld. The region of steady state condition is commonly identified by
visual observation of the welded plates (a good quality surface without
excessive flash formation, voids or cracks), and confirmed by analysis of the
forces (on the longitudinal and vertical direction) on the tool; this is typically
established beyond the first 150 mm of welding.
Fig. 4. Typical chart of tool position, force and torque data measured by the welding machine during FSW.
A consistent metallographic preparation process was applied on each sample
examined and this process made use of the normal met-prep equipment and
consumables. The process consisted of hot mounting, grinding and polishing
on a semi-automatic preparation machine, followed by etching with Nital 2%.
The samples were then positioned on a stage and macro-graphic images of the
weld zone were captured. A more detailed examination of areas of interest
within the weld region identified on the macro-graphs was performed with the
aid of a light optical microscope that involved taking several micrographic
images from each sample being studied.
3.3. Tensile testing
Transverse tensile tests were conducted in parallel to the microstructural
examination study to further support the development of the process envelope,
in that the transverse tensile strength was assessed using a consistent testing
program. This testing program enabled the determination of the yield strength
(YS) and ultimate tensile strength (UTS) that the applied sets of weld
parameters produce on the welded plates, along with the position of the
fracture (parent material or weld metal).
All transverse tensile tests were performed on an Instron 8802 servo-hydraulic
uniaxial tensile testing system in accordance with ISO Standards [11]. Three
samples were sectioned from each weld presented in Table 1 as per the same
specifications in order to improve the repeatability of results (Figure 5). In
addition, three samples were sectioned and tested from the parent material
-40
-20
0
20
40
60
80
100
120
-1000
-500
0
500
1000
1500
2000
2500
0 100 200 300 400 500 600 700
Z F
orc
e (
kN),
Tra
vers
e F
orc
e (
kN)
Z Fo
rce
(kN
),
Trav
ers
e F
orc
e (
kN),
To
rqu
e (
Nm
)
Tim
e (
s), T
rave
rse
Sp
ee
d (
mm
/min
), S
pin
dle
Sp
ee
d
(RP
M),
To
rqu
e (
Nm
)Dis
tan
ce T
rave
lled
(m
m),
Tr
ave
rse
Sp
ee
d (
mm
/min
), S
pin
dle
Sp
ee
d (
RP
M),
Time (s)
Distance Travelled Traverse speed Spindle Speed
Z Position Z Force Traverse Force
Torque
(PM) of three randomly selected welded plates but in parallel to the weld
direction. Thus, the mechanical properties of DH36 in the longitudinal direction
were established and provided values for comparison. The consistent control
method used was an extension rate of 0.5 mm/min up to extension of 1.25 mm,
then 5 mm/min up to fracture.
Fig. 5. Transverse tensile test sample of rectangular cross section (thickness 6 mm).
3.4. Hardness and impact toughness
Micro-hardness measurements were taken across the entire weld zone using a
Mitutoyo hardness tester. The measurements were performed by applying a
load of 200 gf and using a grid spacing of 1 mm (for both x and y directions)
(Figure 6).
Fig. 6. Grid spacing for micro-hardness measurements, Image from weld W021 [x50, Etched].
Charpy impact tests with the standard V-notch were performed according to
ISO Standards [12] in order to evaluate the toughness of the weld region in
relation to the weld parameters used. Samples were sectioned perpendicular to
the weld centreline and transverse to the weld direction (Figure 7). One sample
was sectioned with the notch axis of symmetry on the weld centreline. To
examine the full width of the weld region however, three additional samples
were sectioned towards both sides of the weld in 1.5 mm increments (Figure 8).
In this configuration, each sample with the notch axis at 4.5 mm from the weld
centreline included a small percentage of parent material which was essentially
the same for either side. All other samples fully consisted of thermo-
mechanically stirred material.
Fig. 7. Charpy sample and notch orientation (bottom) relative to welded plate section (top).
Reduced-section samples of 5 mm width were used due to the plate thickness
of 6 mm. In total, three sets of seven tests were performed in room temperature
(~20oC) for each set of parameters to improve the accuracy of the results.
Fig. 8. Typical macro-graph showing the position of the notch axis of symmetry of the seven
Charpy samples examined from each weld.
4. Results and discussion
4.1. Microstructural observations
The following nomenclature is adopted in the present study and the main
regions of the weld zone are illustrated in Figure 9, where:
AD: Advancing side, the side where the rotating FSW tool pushes the metal
towards the weld direction, i.e. forwards. The convention employed for the
entire project is that samples are prepared so that the advancing side is
presented on the left side of all images.
RT: Retreating side, the side where the rotating tool pushes the metal in a
direction opposite to the weld direction, i.e. backwards.
TMAZ: Thermo-mechanically affected zone in which the material has been
thermo-mechanically stirred by the FSW tool.
Weld root: part of TMAZ, around and below the tip of the FSW tool’s pin.
HAZ: Heat affected zone, where the metal has been affected by heat as it
dissipates from the TMAZ, but not mechanically stirred.
PM: Parent material, metal not affected by the process.
Fig. 9. A typical macrograph of the friction stir weld region.
Referring to the classification in Table 1, the group of slow welds presented a
very homogeneous microstructure without any flaws. These exhibited a ferrite
rich microstructure with highly refined grains of random geometry. A very small
content of what is expected to be acicular-shaped bainitic ferrite was found in
W019C (Figure 10), the content of which was seen to steadily increase with
increasing traverse speed and constant rotational speed (200 rpm). This
observation confirms that the cooling rate is increasing with increasing traverse
speed. Additionally, there appears to be a threshold value of approximately 130
mm/min above which this apparently acicular bainitic ferrite microstructure
occurs at a small ratio relative to the above described ferrite rich microstructure.
The parameters of W019A, W019B and W019C resulted in welds being
produced with a large and symmetrical tool footprint on the weld surface. Using
a similar traverse speed of 127 mm/min, Cho et al. [6] observe a predominantly
acicular shaped bainitic ferrite stir zone (TMAZ) microstructure developed as
acicular bainitic ferrites nucleate mainly on the austenite grain boundaries. It is
noted that this is a product of the phase transformation of austenite, in the
supercritical stir zone (central TMAZ), to ferrite at a high cooling rate. Yet,
comparing to the rotational speed of 200 rpm which is used in the 100 – 156
mm/min range of traverse speeds of this study, the rotational speed of 450 rpm
employed by Cho et al. [6] seems high. That is, their traverse speed seems
Weld root
TMAZ
PM
AD RT
PM
rather slow for the considerable heat input deriving from the rotating FSW tool
to dissipate quickly enough, thus producing a low natural cooling rate.
Therefore, it is possible that their resultant microstructure is affected by an
applied forced cooling method, which is not disclosed, and the steel’s different
chemical composition to DH36.
Fig. 10. W019C, microstructure of mid-TMAZ [x1000, Etched].
W021 has been classified in the group of slow weld parameters; contrary to
other welds of this group however, it exhibits an acicular ferrite predominant
microstructure. There is a suggestion that coarse carbides and small, more
randomly mixed bainite-rich regions are observed. On the latter regions, prior
austenite grain boundaries are believed to be present. In all, this group of welds
is expected to have acceptable mechanical properties. As discussed earlier,
another study [4] reports mainly acicular ferrite microstructure in the TMAZ of
Acicular-shaped bainitic ferrite
DH36 welds; thus, it can be concluded that the welding conditions have
produced a comparable cooling rate to the one occurring in weld W021.
Fig. 11. W028, microstructure of mid-TMAZ [x1000, Etched].
In the intermediate group of welds with rotational speed of 400 rpm and as
traverse speed is seen to increase, the microstructure becomes more
heterogeneous, with regions of increasing bainite content (suggesting an
increased cooling rate). However, this heterogeneous microstructure does not
seem to have a significant effect on the mechanical properties (see later). Still,
weld W028 is very homogeneous, with an apparently acicular-shaped bainitic
ferrite microstructure (Figure 11); this would suggest that a good balance of
rotational and traverse speed has been achieved. In addition, prior austenite
grain boundaries can be observed on the light optical microscope, only on the
bainite predominant regions.
Acicular-shaped bainitic ferrite
The friction stir welding process appears to be tolerant to welding speed
variations at the 450 rpm rotational speed. Both welds in this intermediate
group exhibit a homogeneous, fully acicular bainitic ferrite microstructure with
prior austenite grain boundaries. Due to a marginally lower heat input, W029
presents a rather small heat affected zone; weld W030 however seems to be
the product of an excellent intermediate set of parameters (Figure 12).
Fig. 12. W030, microstructure of mid-TMAZ [x1000, Etched].
At the high welding speed of 500 mm/min, the microstructure of all five welds
becomes very heterogeneous, with poorly mixed regions of acicular ferrite and
varying bainite content. The presence of bainite has increased considerably in
this group due to the even higher cooling rate that it is expected to occur. Prior
austenite grain boundaries are easily detected on the regions of bainite
predominant microstructure. Two distinct microstructures co-exist within the
TMAZ of weld W035 (Figure 13); these would be expected to act as stress
concentration regions. Yet, this weld’s parameters seem to have achieved a
Acicular-shaped bainitic ferrite
Prior austenite grain boundaries
good balance of traverse and rotational speed that is translated into satisfactory
tensile behaviour (see later). This could be attributed to the grain refinement
and a suitable ratio of microstructures overtaking the negative effects of
heterogeneity.
Fig. 13. W035, microstructure of mid-TMAZ AD side [x1000, Etched].
Reynolds et al. [3] report that a bainitic and martensitic (therefore acicular by
definition) microstructure has evolved in the TMAZ using the same grade of
steel and a marginally lower traverse speed of 450 mm/min. They [3] conclude
that it is a consequence of the phase transformation of austenite during fast
cooling, and as no ferrite is observed in this region, evidence of FSW elevating
the steel’s temperature above A3. The rotational speed of this weld is not made
clear in order to allow comparisons with the current welds. Still, the formation of
martensite suggests that the cooling rate developed during their welding
process is higher than the rate occurring herein.
Acicular-shaped bainitic ferrite
Prior austenite grain boundaries
Acicular ferrite
The process becomes very sensitive to parameters change, i.e. variations in
rotational speed in this group of welds. Evidence for this can be drawn from the
following two observations:
Welds W035 and W036 only differ by 25 rpm in rotational speed; W036
however is rather unstable, as if steady state conditions have not been
achieved (also see later) and with appreciable carbide precipitation.
In two of the five welds in this group (W032 and W033), incomplete
fusion appears on the advancing side, within which interconnected non-
metallic inclusions are detected. These flaws have a significant length
with secondary paths, and are clearly expected to affect the mechanical
properties of the weld.
There is evidence of minor weld root flaws for most of the examined welds;
regions of insufficient or no welding of the original interface between the two
plates. Moreover, some samples present a poor quality top surface with
fissures introducing non-metallic inclusions in the TMAZ. Nevertheless, both
types of flaws are regarded as processing features and have been reported to
TWI for further investigation.
As a general note, no surface voids or cracks or others flaws are detected
(except those mentioned above for welds W032 and W033) in the bulk of the
weld zone (TMAZ), and especially in the top advancing side where the shear
forces on the metal are expected to be the highest. Hence, high speed friction
stir welding of steel grade DH36 is feasible; the process is tolerant to parameter
variation in low and intermediate speeds but less tolerant at the highest speed
(500 mm/min).
4.2. Mechanical property assessment
Stress vs. strain charts were plotted for each transverse tensile test to evaluate
the consistency of the generated data and to determine the yield strength and
ultimate tensile strength (UTS) of each sample. The calculated mechanical
properties are outlined in Table 3.
The samples of fifteen out of the nineteen welds fractured in the parent material
rather than the weld (all three samples per weld). This suggests that these
fifteen sets of weld parameters produced welds of higher strength than the
parent material. All samples prepared from the fourteen slow and intermediate
welds (as classified in Table 1) fractured in the parent material. Even welds
which consist of high acicular bainitic ferrite phase ratio and therefore could be
expected to show brittle behaviour proved to be stronger than the parent
material. Identical tensile behaviour is reported in other studies [3,4], where all
transverse tensile samples produced from welds of similar sets of parameters
fractured in the parent material, well away from the HAZ. Equally, welds with a
very heterogeneous microstructure (e.g. W031), i.e. containing stress
concentration regions which may act as crack initiation points, or welds with
minor flaws such as weld root flaw or fissures on the top surface fractured in
the parent material.
Table 3
Summary of DH36 FSW transverse tensile tests results.
Weld Transverse tensile testing at Room Temperature
Reference number
Rotational Speed (rpm)
Traverse Speed (mm/min)
Yield Strength
(0.2%), MPa
UTS, MPa
Region of fracture
Fracture mode
W019A 200 100
395 530
PM Ductile 408 540
398 536
W019B 200
110
390 530
PM Ductile 394 536
386 535
W019C 200 120
401 536
PM Ductile 396 529
392 527
W020A 200 130
406 535
PM Ductile 407 536
401 533
W020B 200 143
398 521
PM Ductile 400 527
409 529
W020C 200 156
405 531
PM Ductile 396 523
397 525
W021 400 200
320 512
PM Ductile 380 516
391 528
W022 400 250
384 525
PM Ductile 384 523
381 519
W023 400 275
383 522
PM Ductile 384 521
378 525
W027 400 325
387 526
PM Ductile 387 525
388 524
W028 400 375
390 528
PM Ductile 390 530
392 525
W029 450 400
397 540
PM Ductile 396 534
393 532
W030 450 350
399 539
PM Ductile 405 544
396 534
W031 300 250
382 519
PM Ductile 385 514
386 522
W032 600 500
383 457 Weld, AD
side Brittle 384 458
400 488
W033 650 500 412 526 Weld, AD
side Brittle
408 563
417 534
W034 575 500
423 433 Weld, AD
side Brittle 429 462
442 480
W035 700 500
382 519
PM Ductile 388 523
401 546
W036 675 500
397 514 Weld, AD side
Brittle
390 458 Brittle
394 532 PM Ductile
PM – 1
Longitudinal direction
339 521
PM Ductile PM – 2 340 525
PM – 3 340 526
The group of fast welds (500 mm/min) demonstrated a different behaviour. All
samples from welds W032, W033 and W034 fractured in the weld, and
specifically in a brittle-like manner on the outer boundary of the advancing side.
This fracture region appears to correspond to the top AD TMAZ region where
non-metallic inclusions were found to be interconnected in an incomplete fusion
characteristic. This feature was observed in welds W032 and W033.
Two samples from weld W036 fractured inside the weld (brittle fracture on outer
AD side) and one fractured in the parent material in a ductile manner (Figure
14). In fact, there is a noteworthy difference in the fracture faces of the first two
samples; one sample’s fracture face has the FSW tool’s features imprinted on it
(on the mid-AD TMAZ, Figure 14a). This denotes that the metal was
mechanically worked (similar to forging) by the FSW tool in this region rather
than thermo-mechanically stirred, perhaps due to lower heat input. The second
sample presents markings consistent with lower (than required) temperature
material flow patterns, suggesting lower heat input in this region (Figure 14b).
Figure 14c presents the third sample examined from weld W036, with a typical
ductile fracture. Such observations indicate that this set of parameters is on a
threshold value producing welds where steady state conditions have not been
reached (an “unstable” weld).
W035 is the only one of the fast weld category that fractured in the parent
material with a typically ductile fracture. Although a very heterogeneous
microstructure (with possible stress concentration regions associated with this
heterogeneity) has been observed on the optical microscope, all three samples
proved to be stronger than the parent material. It would appear that this set of
parameters achieves a good balance of rotational speed (sufficient heat input
for proper thermo-mechanical stirring) and traverse speed (affecting the cooling
rate which governs the ratio of phases) to produce acceptable quality welds.
a
b
c
Fig. 14. Three samples of W036 exhibiting different fracture behaviour.
On the hardness of the weld region and for reporting purposes, two welds from
each group of slow, intermediate and fast welds will be discussed herein. In
addition, all hardness values presented in Figure 15 refer to the top-TMAZ of
each weld sample, i.e. 1 mm below the top surface (Figure 6). In the slow and
intermediate welds (Figures 15a & 15b), the hardness of the weld is higher than
that of the parent material but not at levels which can cause concern
considering their tensile test results. The hardness of each weld seems to be
symmetrical across the weld region, with a relatively even distribution from
FSW tool markings
Material flow markings
Characteristic “necking” of ductile fracture
advancing to retreating side of the TMAZ. Further, the hardness of the weld
region in both groups is seen to increase with increasing cooling rate of the
weld. In absolute terms, both intermediate welds have produced higher
hardness values compared to the group of slow welds, as a result of their lower
heat input and higher cooling rate hence increasing bainite content.
0
50
100
150
200
250
300
350
-6 -4 -2 0 2 4 6
Ha
rdn
es
s, H
V
Distance from weld centreline, mm
(a) Slow welds, top-TMAZ (5mm)
W019A
W021
PMRT
0
50
100
150
200
250
300
350
400
450
-6 -4 -2 0 2 4 6
Hard
ne
ss
, H
V
Distance from weld centreline, mm
(b) Intermediate welds, top-TMAZ (5mm)
W028
W030
PM
AD
AD RT
Fig. 15. Micro-hardness distribution of the weld region for three groups of weld parameters.
In the group of fast welds (Figure 15c), W035 exhibits a high but still
reasonable range of hardness values with minor variations across the weld
region. This may partly explain the excellent behaviour of this specific weld’s
samples in the transverse tensile tests. As expected due to its higher cooling
rate, W036 is seen to produce higher hardness values particularly on the
advancing side, and with more pronounced variations. This reflects the
microstructural heterogeneity of this weld with region of varying bainite
concentration, thus seems to suggest that different regions of the welded
material, as it is stirred from the advancing to the retreating side, are
experiencing different cooling rates. The very high hardness value on the outer
advancing side should correlate with the same weld’s tensile samples fracturing
at the same position. In absolute terms again, the fast welds have produced
similar hardness levels to the intermediate group suggesting similar cooling
rates.
Such variations in the hardness distribution within the TMAZ of most welds
currently examined are often reported by other researchers. Ghosh et al. [5]
note that the significant variations of hardness within the weld zone of all
welding parameters examined occur due to the heterogeneity of the resultant
microstructure, from weld nugget to heat affected zone. Further, the differences
in peak hardness between weld nugget regions depend on the weld
parameters used hence the subsequent cooling rate of each weld. Cho et al. [6]
argue that the substantially higher hardness found in the stir zone (central
TMAZ) is caused by the acicular shaped bainitic ferrite microstructure. Similar
microstructure is observed in most of the intermediate and high speed welds of
this study but with smaller variations in the hardness of the TMAZ, suggesting a
reasonably homogeneous microstructural distribution.
0
50
100
150
200
250
300
350
400
450
-6 -4 -2 0 2 4 6
Ha
rdn
es
s, H
V
Distance from weld centreline, mm
(c) Fast welds, top-TMAZ (5mm)
W035
W036
PM
RT AD
Fig. 16. Impact toughness distribution in the weld region for six sets of weld parameters at 20
oC.
The impact toughness data presented in Figure 16 have been normalised to a
10 x 10 mm equivalent by a scaling factor of 3/2, as discussed by McPherson
et al. [4]. The impact toughness of most welds at 20oC appears to have been
reduced compared to the parent material (Figure 16). The slow welds follow a
similar pattern of peak values on the weld centreline and lowest values nearing
both sides of the TMAZ. It appears that the presence of acicular ferrite in W021
has a detrimental effect on the impact toughness when compared to the ferrite-
rich (with refined grains of random geometry) microstructure in W019A. All
intermediate and fast welds exhibit a similar trend in impact toughness
distribution, with a peak observed in the inner TMAZ of the advancing side and
gradual decrease towards the outer boundaries of the TMAZ on both sides. The
homogeneous acicular bainitic ferrite microstructure observed in W028 has
produced marginally higher impact toughness in the AD TMAZ than the
heterogeneous microstructure of W035. The latter is characterised as having
two distinct structures, acicular bainitic ferrite and acicular ferrite; as with weld
W021 discussed above, the presence of acicular ferrite seems to be
responsible for the slightly reduced impact toughness of W035. Considerable
variation between the impact toughness of the advancing and retreating sides
is apparent in the intermediate and fast welds. It should be highlighted that two
0
20
40
60
80
100
120
140
-5 -4 -3 -2 -1 0 1 2 3 4 5
Ab
so
rbe
d e
ne
rgy,
J a
t 2
0o
C
Distance from weld centreline, mm
Impact toughness, 6 indicative welds
W019A W021
W028 W030
W035 W036
PM, weld direction
RT AD
sets of parameters which were previously identified as being well balanced,
W028 and W035, overmatch the parent plate’s impact toughness in the inner
AD TMAZ.
The data in Figure 16 reveal that increasing the traverse speed from 100
mm/min to 375 mm/min and 500 mm/min delivers a significant effect of
improving the impact toughness on the advancing side, although having a
lesser improvement on the retreating side. The impact toughness in both sides
of the outer TMAZ however is seen to decrease to some extent with the
intermediate and fast traverse speed. Still, the improvement in impact
toughness offers a level of confidence in increasing the FSW traverse speed.
5. Conclusions
1. A comprehensive study, consisting of 19 friction stir welds of 2000 x 200 x 6
mm DH36 steel plates has generated an extensive data set to account for a
wide range of typical and atypical process parameters. This has resulted in
the development of a preliminary FSW parameter envelope based on the
outcomes of microstructural evaluation and mechanical testing.
2. The welding (traverse) speed on DH36 steel has been increased to a great
extent compared to the conventionally used speed of 100 mm/min. This
represents considerable potential economic advantages in terms of the
competitiveness of FSW to conventional welding processes.
3. An initial understanding of the relationship between weld parameters and
the effect of these on weld microstructures and mechanical properties has
been established. From this, it is evident that FSW generates a very
complex metallurgical system in which the slow speeds result in a very
refined, ferrite rich microstructure, the intermediate speeds produce
predominantly acicular bainitic ferrites, and the fast speeds result in a
heterogeneous microstructure with distinct regions of acicular ferrite and
acicular bainitic ferrite.
4. A number of weld parameters have been identified which may produce fast
(in the region of 400 mm/min) or very fast (potentially up to 500 mm/min)
welds of acceptable quality that are highly competitive to conventional
fusion welding techniques.
5. Increasing welding (traverse) speed is seen to improve the impact
toughness of the weld without compromising strength and hardness.
Subsequent work will include a further refinement of the high speed weld
parameters identified in this characterisation study and a comprehensive
fatigue testing program.
Acknowledgements
The authors gratefully acknowledge the financial support of the European
Union which has funded this work as part of the Collaborative Research Project
HILDA (High Integrity Low Distortion Assembly) through the Seventh
Framework Programme (SCP2-GA-2012-314534-HILDA).
References
[1] W.M. Thomas, E.D. Nicholas, J.C. Needham, M.G. Murch, P. Temple-
Smith, and C.J. Dawes: GB Patent No. 9125978.8, International Patent No.
PCT/GB92/02203.
[2] W.M. Thomas, P.L. Threadgill, and E.D. Nicholas: Sci. Technol. Weld.
Join., 1999, vol. 4 (6), pp. 365-72.
[3] A.P. Reynolds, W. Tang, M. Posada, and J. Deloach: Sci. Technol. Weld.
Join., 2003, vol. 8 (6), pp. 455-60.
[4] N.A. McPherson, A.M. Galloway, S.R. Cater, and S.J. Hambling: Sci.
Technol. Weld. Join., 2013, vol. 18 (5), pp. 441-50.
[5] M. Ghosh, K. Kumar, and R.S. Mishra: Metall. Mater. Trans. A, 2012, vol.
43A, pp. 1966-75.
[6] H.H. Cho, S.H. Kang, S.H. Kim, K.H. Oh, H.J. Kim, W.S. Chang, and H.N.
Han: Mater. Des., 2012, vol. 34, pp. 258-67.
[7] S. Kumar, S. Kumar, and A. Kumar: Proc. Inst. Mech. Eng., Part C, 2012,
vol. 227 (6), pp. 1150-63.
[8] M. Ghosh, K. Kumar, S.V. Kailas, and A.K. Ray: Mater. Des., 2010, vol. 31,
pp. 3033-37.
[9] Z.Y. Ma: Metall. Mater. Trans. A, 2008, vol. 39 (3), pp. 642–58.
[10] H. Liu, M. Maeda, H. Fuji, and K. Nogi: J. Mater. Sci. Lett., 2003, vol. 22
(1), pp. 41–3.
[11] British Standards Institution: BS EN ISO 6892-1:2009, London, U.K.,
August 2009.
[12] British Standards Institution: BS EN ISO 148-1:2010, London, U.K.,
November 2010.